11671
J. Am. Chem. Soc. 1993,115, 11671-1 1675
Engineered Biosynthesis of Novel Polyketides: Manipulation and Analysis of an Aromatic Polyketide Synthase with Unproven Catalytic Specificities Robert McDaniel,'" Susanne.Ebert-Kbosla,'* David A. Hopwood,lb and Chaitan Khosla',l* Contributionfrom the Department of Chemical Engineering, Stanford University, Stanford, California 94305-5025, and Department of Genetics, John Innes Institute, John Innes Centre, Norwich NR4 7UH, U.K. Received July 15. 1993@
Abstract:Genes for a putative aromatic polyketide synthase (PKS) of unproven catalytic specificity from Streptomyces roseofuluus,which produces both frenolicin B (1)and nanaomycin A (2), were functionally expressed in Streptomyces coelicolor A3(2) using a heterologous expression system in combination with components of the actinorhodin ( 3 ) PKS gene cluster. Spectroscopic characterization and isotopic labeling experiments showed that the primary product of the recombinantestrainwas a novel tricyclic molecule (6), derived from a nonaketide backbone that undergoes a single ketoreduction. In addition, two octaketides (4 and 7) were produced by this PKS, suggesting that the chain length determining factor (CLF) of this PKS has a relaxed specificity and providing strong support for the supposition that the cloned genes encoded the PKS for frenolicin and nanaomycin biosynthesis. The regiospecificity of ketoreduction of the putative backbone of one of the octaketides (7)argues in favor of a biosynthetic model in which ketoreduction occurs only after the complete polyketide chain has been synthesized. Construction and analysis of hybrid PKSs containing different subunits from the actinorhodin and the above PKS confirmed earlier conclusions regarding the protein determinants of chain length, ketoreduction, and cyclization specificities.
Polyketides are a large, structurally diverse family of natural products possessing a broad range of biological activities,including antibiotic and pharmacological properties. Biosynthetic and molecular genetic studies have clearly demonstrated that polyketide synthases (PKSs) are structurally and mechanistically related to each other and to fatty acid synthases (FASS).~The PKSs are multifunctionalenzymes that catalyze repeated (decarboxylative) Claisen condensations between acylthioesters (usually acetyl, propionyl, malonyl, or methylmalonyl). Following each condensation, they introduce structural variability into the product by catalyzing all, part, or none of a reductive cycle comprising a ketoreduction, dehydration, and enoylreduction on the b-keto group of the growing polyketide chain. After the carbon chain has grown to a length characteristic of each specific product, it is released from the synthase by thiolysis or acyl transfer. Recently, we described a generally applicable strategy for the biosynthesis of novel polyketides by the functional expression of recombinant PKSs in a speciallyconstructed expression system.' Analysis of the structures of the molecules revealed key features of the mechanisms by which the PKSs for polycyclic aromatic compounds control their catalytic specificities. It was shown that (i) the chain length is, at least in part, dictated by a protein called the chain length determining factor (CLF; Figure l), (ii) while the association of some heterologous ketosynthase/acyltransferase (KS/AT; Figure l)-CLF pairs gave rise to functional PKSs, other pairs were nonfunctional, (iii) acyl carrier proteins (ACPs; Figure 1) could be interchangedamong differentsynthases without any effect on product structure, (iv) a given ketoreductase (KR; Figure 1) recognized and reduced polyketide chains of Abstract published in Advance ACS Abstracts. November 15, 1993. (1) (a) Stanford University. (b) John Innes Centre. (2) (a) Wakil, S. J. Biochemistry 1989,28,4523. (b) Hopwood, D. A.; Sherman, D. H. Annu. Rev. Genet. 1990, 24, 37. (c) OHagan, D. The PolyketideMetabolites;Ellis Horwood: Chichester,U. K., 1991. (d) Robinson, J. Philos. Trans. R. SOC.London, B 1991, 332, 107. (e) Hopwood, D. A.; Khosla, C. Ciba Found. Symp. 1992,171,88. (f) Hutchinson, C. R.; Decker, H.; Motamedi,H.; Shen, B.; Summers,R. G.;Wendt-Pienkowski,E.; Wessel, W. L. Actinomycetology 1992, 6, 49. (g) Katz, L.; Donadio, S. Ann. Rev. Microbiol. 1993, 47, 875. (3) McDaniel, R.; Ebert-Khosla,S.; Hopwood, D. A.; Khosla, C. Science in press.
ketosynthase
0 acyltransferase chain length determining factor
acyi carrier protein ketoreductase cyclase
Figure 1. The act andfren PKS gene clusters. Each PKS includes a ketosynthase/acyltransferase(KS/AT, ORFl), a chain length determining factor (CLF, ORF2), an acyl carrier protein (ACP, ORF3), and a ketoreductase (KR). The AT presumably tranfers the starter unit from CoA to the KS, which catalyzes the Claisen condensation between the starter (or growing polyketide) acylthioesterand the extender thioester on the ACP. The KR reducescarbonylsof the polyketidebackbone. The gene clusters also code for a cyclase, involved in cyclization of the nascent polyketide backbone. The genetic homology depicted here is conserved throughout most actinomycete aromatic PKSs.
different lengths, (v) regardless of the chain length, this ketoreductasecatalyzed a singleketoreductionat the C-9 position, counting from the carboxyl end, (vi) if this ketoreduction was allowed to occur, the regiospecificity of the first cyclization was constant relative to the position of ketoreduction, regardless of the chain length, and (vii) a specificcyclase (Figure 1) responsible for the second cyclization reaction could not cyclize substrates of altered chain lengths. Here, we extend the earlier approach' by generating recombinant PKSs incorporating components from a PKS gene set of unproven (though conjectured) specificity, and use the structures of the novel polyketides to confirm the role of the unknown PKS. The genes encoding such PKSs can readily be cloned from actinomycete species by taking advantage of the high degree of
OOO2-7863/93/1515-11671$04.O0/0 0 1993 American Chemical Society
11672 J. Am. Chem. SOC., Vol. 115, No. 25, 1993 sequence similarity between members of different PKS subg r o u p ~ . ~Using - ~ this expression system, it should therefore be possible to generate a wide variety of novel polyketide backbone chains by constructing a library of recombinant PKSs with components cloned from different polyketide producers. Isotope labeling and N M R analysis of the products of the recombinant PKS described here have allowed us to address two key questions that arose in our earlier work3and that of (i) does ketoreduction of polycyclic aromatic polyketide backbones occur in the ,!?-position, or does it occur after the complete carbon backbone is synthesized, and (ii) do cyclases discriminate between polyketide chains of different lengths, or is their inability to catalyze the cyclization of polyketide chains of altered lengths the result of lack of recognition of CLFs having different chain length specificities? We have also identified the first example of a PKS with a relaxed chain length specificity that allows it to produce both octaketide and nonaketide products.
McDaniel et al.
HO
1
S-E Frenolicin B (1)
a 0
HO
0
0
OH
0
i
S E-)--
Hi$& /$
COOH
0 Nanaomycin A (1)
OH
0
OH
0
RC!Sults HO
Streptomyces roseofulvus produces both frenolicin B (1)8and nanaomycin A (Z).9 The carbon chain backbones of these aromatic polyketide products have been hypothesized to be derived from nine and eight acetates, respectively,2ealthough in vivo acetate labeling studies have not been carried out on this strain. Furthermore, it is presumed that both carbon chains undergo a 8-ketoreduction at the C-9 position and that the frenolicin HO backbone also undergoes a complete cycle of ,!?-ketoreduction, dehydration, and enoylreduction a t the C-17 position. A 10-kb DNA fragment (referred to as thefren fragment hereafter) was cloned from a genomic library of S. roseofulvus5 using DNA encoding the KS/AT and KR components of the actinorhodin (3) (act) PKS ofS. coelicolorA3(2) as a hybridization p r ~ b e . ~ ~ J ~ The act PKS catalyzes the biosynthesis of the octaketide precursor of actinorhodin, another aromatic polyketide, whose backbone is derived from eight acetates with a single ketoreduction at the C-9 position (Le. identical to that of nanaomycin).lI DNA sequencing of thefren fragment5revealed the existence of genes with a high degree of identity to those encoding the genes from the act cluster involved in carbon chain formation and cyclization; these are actl-ORF1, -2, and -3 genes that encode a putative bifunctional ketosynthase/acyltransferase (KS/AT), a chain length determining factor (CLF), and an acyl carrier protein (ACP), respectively; a ketoreductase (KR); and a cyclase (Figure 1). The host-vector system used for expression and analysis of recombinant PKSs in this work is identical to that reported earlier;3 the vector was pRMS and the host was S . coelicolor CH999 (a mutant from which the act cluster has been precisely deleted). ~
COOH
Actinorhodin (3)
fi--- &$y OH
0
I
OH
SE
0
B
COOH 3,l-dihydroxy-lmethylanthraquinone-2carboxylic acid (4)
H
Alcesaponarin II (5)
HO
0
8
8
4
2
RM18 (6)
0
HO
RM18b (7)
~~~
(4) (a) Malpartida, F.; Hallam, S. E.; Kieser, H. M.; Motamedi, H.; Hutchinson, C. R.; Butler, M. J.; Sugden, D. A.; Warren, M.; McKillop, C.; "0, /20 Bailey, C. R.; Humphreys, G. 0.;Hopwood, D. A. Nature 1987, 325, 818. (b) Arrowsmith,T. J.; Malpartida, F.; Sherman, D. H.; Birch, A.; Hopwood, D. A.; Robinson,J. A. Mol. Gen. Genet. 1992,234,254. (c) Bergh, S.; Uhlen, M. Gene 1992,117, 13 1. (d) Legouill, C.; Desmarais, D.; Dery, C. V. Mol. Gen. Genet. 1993, in press. (5) Bibb, M. J.; Sherman, D. H.; Omura, S.; Hopwood, D. A. Submitted. ( 6 ) Bartel, P. L.; Zhu, C. B.; Lampel, J. S.; Dosch, D. C.;Connors, N. C.; Strohl, W. R.; Beale, J. M.; Floss, H. G. J . Bacteriol. 1990, 172, 4816. 8 8 4 2 (7) (a) Zhang, H. L.; He, X. G.; Adefarati, A,; Gallucci, J.; Cole, S. P.; Beale, J. M.; Keller, P. J.; Chang, C. J.; Floss, H. G. J. Org. Chem. 1990.55, RM20 (8) 1682. (b)Shennan,D.H.;Bibb,M.J.;Simpson,T.J.;Johnson,D.;Malpartida, Figure 2. Polyketides (right) and deduced backbones (left) synthesized F.; Fernandez-Moreno,M.; Martinez, E.; Hutchinson, C. R. Hopwood,D. A. Tetrahedron 1991, 47, 6029. (c) Summers, R. G.; Wendt-Pienkowski,E.; by PKSs. Motamedi, H.; Hutchinson, C. R. J . Bacteriol. 1992, 174, 1810. The host was transformed with a pRMS derivative containing (8) Iwai, Y.;Kora, A,; Takahashi, Y.;Hayashi, T.; Awaya, J.; Masuma, R.; Oiwa, R.; Omura, S. J . Antibiot. 1978, 31, 959. the fren PKS genes that are presumably responsible for the (9) Tsuzuki, K.;Iwai, Y.; Omura, S.;Shimizu, H.;Kitajima,N. J. Antibiot. biosynthesis of a polyketide backbone (Le. ORFs 1-3) (Table I).
1986, 39, 1343. (10) (a) Malpartida, F.; Hopwood, D. A. Nature 1984, 309, 462. (b) Malpartida, F.; Hopwood, D. A. Mol. Gen. Genet. 1986, 205, 66. (c)
Fernandez-Moreno, M. A,; Martinez, E.; Boto, L.; Hopwood, D. A,; Malpartida, F. J. Bioi. Chem. 1992, 267, 19278. (11) Gorst-Allman, C. P.; Rudd, B. A. M.; Chang, C. J.; Floss, H. G. J . Org. Chem. 1981, 46,455.
In addition, plasmids carrying all possible combinations of act andfren ORFs 1-3 were also constructed and used to transform CH999 (Table I). The genes encoding the act KR and the act cyclase were also present on each of these genetic constructs. On the basis of results from similar experiments with act and tcm
J. Am. Chem. Soc., Vol. 1 1 5, No. 25, 1993 11673
Engineered Biosynthesis of Novel Polyketides Table I. Summary of the Major Polyketide Products Isolated from Recombinant Strains with Various Combinations of Act and Fren ORFs 1-3 Genesa plasmid pRM5 pRM8 pRM 13 pRM23 pRM 18 pRM32 pRM33 pRM34
ORFl (=/AT) act fren act act fren fren act fren
ORF2 (CLF)
ORF3 (ACP)
act act fren act fren act fren fren
act act act fren act fren fren fren
major products 475 4,s
NP 495 4,5,6,1
NP NP 4,5,6,7
NP, no product.
u-.
NOE Interactions
HETCOR Summary
Figure 3. Summary of NOE and HETCOR (HMBC) data from RM18 (6).
(tetracenomycin) PKSS,~ it was expected that the act KR would be able to reduce the products of all functional recombinant PKSs, whereas the ability of the act cyclase to catalyze the second cyclization would depend upon the chain length of the product of the fren PKS. The results summarized in Table I indicate that most of the transformants expressed functional PKSs, as assayed by their ability to produce aromatic polyketides. Structural analysis of the major products revealed that the producer strains could be grouped into two categories: those that synthesized compound 4 (together with a smaller amount of its decarboxylated side product (5)) and those that synthesized a mixture of compounds 4, 6, and 7 in a roughly 1:2:2 ratio. (Small amounts of 5 were also found in all strains producing 4.) Compounds 4 (3,8dihydroxy-1-methylanthraquinone-2-carboxylic acid)12 and 5 (aloesaponarin 11)6had been observed before as natural products and were the metabolites produced by a PKS consisting entirely of act subunit^.^ Compounds 6 and 7 (designated RM18 and RM 18b, respectively) however are novel structures whose chemical synthesis or isolation as natural products has not been reported previously. Thestructuresof 6and 7 were elucidatedthrough a combination of mass spectroscopy, NMR spectroscopy, and isotope labeling experiments. The IH and 13Cspectral assignments are shown in Table I1 along with I3C-l3C coupling constants for 6 obtained through sodium [ 1,2-13C2]acetatefeedingexperiments(described below). Unequivocal assignments for compound 6 were established with 1D nuclear Overhauser effect (NOE) and long-range heteronuclear multiple bond correlation (HMBC) studies (Figure 3). Deuterium exchange confirmed the presence of hydroxyls at C- 15 of compound 6 and C- 13 of compound 7. Field desorption mass spectrometry (FD-MS) of 6 and 7 revealed molecular weights of 282 and 240, respectively, consistent with C17H1404MW (282.2952) and C15H1203MW (240.2580). Earlier studies showed that the polyketide backbone of 56(and by inference, 43) is derived from iterative condensations of eight acetate residues with a single ketoreduction at C-9 (Figure 2). Nanaomycin (2) probably arises from an identical carbon chain backbone (Figure 2). Therefore, it is very likely that nanaomycin (12) (a) Cameron, D. W.; Deutscher, D. J.; Feutrill, G. I.; Griffiths, P. G.
Ausr. J . Chem. 1981, 34, 2401. (b) Krupa, J.; Lessmann, H.; Lacher, H. Liebigs Ann. Chem. 1989, 7, 699.
is a product of the fren PKS genes in S. roseofulvus. Regiospecificity of the first cyclization leading to the formation of 4 is constant relative to the position of the ketored~ction,~ whereas that of the second cyclization is controlled by the act cyclase.7b In order to trace the carbon chain backbone of RM18 (6), in vivo feeding experiments using [ 1,2-13C2]acetatewere performed onCH999/pRM18, followed byNMRanalysisoflabeledRM18. The I3Ccouplingdata (summarized in Table 11) indicate that the polyketide backbone of RM18 is derived from nine acetate residues, followed by a terminal decarboxylation (the C-2 l3C resonance appears as an enhanced singlet), which presumably occurs nonenzymatically (Figure 2). Furthermore, the absence of a hydroxyl group at the C-9 position suggests that a ketoreduction occurs at this carbon. Since such a reduction on an 18-carbon chain would be expected to occur in the putative frenolicin (1) backbone, the results suggest that, in addition to synthesizingnanaomycin, thefren PKS genes are responsiblefor the biosynthesis of frenolicin in S. roseofulvus. This, to our knowledge, represents the first unambiguous case of a PKS with relaxed chain length specificity. However, unlike the putative backboneof frenolicin, the C- 17 carbonyl of RM 18 is not reduced. This could either reflect the absence from pRM18 of a specific ketoreductase, dehydratase, and enoylreductase (which might be present in the “fren”gene cluster in S.roseofulvus), or it could reflect a different origin for carbons 15-18 in frenolicin (for example, butyrate could be used as the starter unit for frenolicin biosynthesis). Regiospecificityof the first cyclizationleading to the formation of RM18 is constant relative to the position of the ketoreduction; however, the second cyclization occurs differently from that in 1or 4 and is similar to the cyclization pattern observed in RM20 (8), a decaketide produced by the tcm PKS.3 Therefore, as in the case of RM20, it could be argued that the act cyclase cannot catalyze the second cyclization of the RM18 precursor and that its subsequent cyclizations, which presumably occur nonenzymatically, are dictated by temporal differences in release of different portions of the nascent polyketidechain into an aqueous environment. In view of the ability of CH999/pRM18 (and CH999/pRM34) to produce 4, one can rule out the possibility that the cyclase cannot associate with the fren PKS (KS/AT, CLF, and ACP). A more likely explanation is that the act cyclase cannot recognize substrates of altered chain lengths. This would also be consistent with the putative biosynthetic scheme for ~ ~ 2 0 . 3 Although isotopelabeling analysis on RM18b (7) has not been performed, upon the basis of a comparison with the carbon chain precursors of 4 and 6, it is likely to be derived from eight acetates, followed by a terminal decarboxylation (Figure 2). (The alternative explanation, namely that RM18b results from deacetylation of RM18, is unlikely, particularly in light of the observation that RM2O-producing cultures do not contain an equivalent product.) If this is true, then ketoreduction must have occurred at the C-7 position (instead of the more commonly observed C-9 position) in the polyketide chain precursor of RM 18b (Figure 2). The simplestexplanation for this unusual observation is that the ketoreductionoccursfollowingsynthesisof the complete polyketide chain, as opposed to when the target carbonyl is in the &position (which is how ketoreductions occur in the case of macrolides;for example, see ref 13). Thus, given the observation that the fren PKS has a relaxed chain length specificity (see above), the somewhat loose fit of the shorter polyketide within a putative chain length determining pocket would result in occasional “slippage” of the full-length PKS product, giving rise to a stochastic mixture of C-7 and C-9 reduced chains (Figure 4). The same hypothesis has also been proposed earlier upon the basis of the observation that ketoreduction always occurs at a (13) (a) Cane, D. E.; Yang, C. J . Am. Chem. SOC.1987,109, 1255. (b) Yue, S.; Duncan, J. S.; Yamamoto, Y.; Hutchinson, C. R. J . Am. Chem. SOC. 1987,109, 1253.
11674 J. Am. Chem. Soc., Vol. 115, No. 25, 1993
McDaniel et al.
Table II. lH (400 MHz) and 13C (100 MHz) NMR Data from RM18 (6) and RM18b (7)
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
2.2 (s, 3H)
NCb 37.7 36.9 77.2 77.4 61.9 62.1 58.9 59.2 56.0 55.7 74.7 74.9 69.6 69.3 41.9 41.7
29.6 203.7 47.0 149.6 106.7 129.1 1 14.4 130.1 120.6 132.7 116.7 155.6 98.4 158.8 113.6 201.7 32.4
3.6 (s, 2H) 6.2 (s, 1H) 6.7 (d, 7.2, 1H) 7.3 (dd, 8.4,7.4, 1H) 7.6 (d, 8.9, 1H)
6.4 (s, 1H) 11.2 (s, 1 0 H ) 2.5 (s, 3H)
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
18.8 152.3 104.0 130.0 113.5 130.1 120.1 132.8 116.6 155.9 98.2 159.1 113.8 201.7 32.4
2.1 (s, 3H) 6.1 (s, 1H) 6.7 (d, 7.0) 7.3 (dd, 7.1, 8.7, 1H) 7.6 (d, 8.6, 1H)
6.4 (s, 1H) 11.2 (s, 1 0 H ) 2.5 (s, 3H)
Carbons are labeled according to their number in the polyketide backbone (Figure 2). NC, not coupled.
origin that appear to be nonproducingwere discussed el~ewhere.~ Two other hybrid PKSs not reported in Table I are also nonproducing. These include a fren ORFl /tcm ORF2 hybrid and a tcm ORFllfren ORF2 hybrid, with the act ACP gene present on both constructs.
-
RM18b
I=
"slip"
-
4
Figure 4. Relaxedspecificityoffren PKS. This schematic representation illustrates proposed orientations of the different polyketide backbones produced by the recombinant fren PKSs and the proposed "slippage" mechanism that explains the aberrent reduction of RM18b (7). See text for further details.
constant distance from the end of the chain (i.e. the carboxyl group), as opposed to the beginning of the chain (i.e. the methyl group)? With regard to the cyclization of RM18b, the first cyclization is again constant relative to the position of the ketoreduction, while the resulting product, albeit an octaketide, is presumably not recognized by the act cyclase. A comparison of the product profiles of the hybrid synthases reported in Table I with analogous hybrids between act and tcm PKS components supports the hypothesis that the ORF2 product is the chain length determining factor (CLF).3 Possible explanations for the hybrid PKSs with ORFl and ORF2 of different
Discussion The isolation of mutant enzymes with relaxed substrate specificities has provided valuable mechanistic insights into enzyme function on many previous occasions.l4 Our deduction that thefren PKS catalyzes the biosynthesis of polyketides of two different chain lengths is the first example of a naturally occuring PKS with a relaxed chain length specificity. Such examples, although relatively rare, are suspected to exist in other classes of PKSs as well.I5 Although a better understanding of the molecular basis of this phenomenon must await further analysis of thefren PKS, the results presented here prove that ketoreductions of unreduced chains can occur at positions other than C-9 and strongly support an earlier hypothesis that ketoreduction of aromatic polyketides occurs after assembly of the complete polyketide chain? The simplestexplanation for the dual specificity of thefren PKS is that it has a binding pocket in which the smaller nascent polyketide backbone can only fit loosely, thus leading to chain slipping prior to ketoreduction. The results reported here also support many of the conclusions drawn from our studies with act and tcm hybrid PKSS;~(i) the ORF2 product (CLF) of aromatic PKSs dictates chain length, at least in part, (ii) the regiochemistry of the first cyclization is constant relative to the position of ketoreduction when it occurs, and (iii) the act cyclase does not recognize polyketide chains longer than eight acetates. Furthermore, examination of the structure of compound 7 suggests that, since the cyclase fails to recognizeeven octaketideswith aberrant ketoreductions,substrate specificity must be dictated by the methyl end of the carbon chain. Finally, the ability to generate novel polyketide structures through genetic engineering of PKSs has been further illustrated by the generation of RM18 (6) and RM18b (7). Given the tremendously diverse catalytic specificitiesexhibited by naturally occuring PKSs, heterologous expression of uncharacterized or poorly characterized PKSs could also be exploited as a very useful synthetic and analytical tool. For example, it should be possible to construct a library of randomly cloned ORFs 1 and 2 homologuesfrom a collection of actinomycetes,which would then (14) (a) Wilks, H. M.; Halsall, D. J.; Atkinson, T.; Chia, W. N.; Clarke, A. R.; Holbrook, J. J. Biochemistry 1990, 29, 8587. (b) Bone, R.; Silen, J. L.; Agard, D. A. Nature 1989, 339, 191. (15) Cane, D. E.; Lambalot, R. H.; Prabhakaran, P. C.;Ott, W. R. J. Am. Chem. SOC.1993,115,522.
Engineered Biosynthesis of Novel Polyketides be screened for their ability to catalyze novel aromatic polyketide production as part of hybrid synthases. One could also attempt to correctly cyclize longer polyketides by replacing the uct cyclase gene with a homologue from PKS gene clusters which produce chains of the correct lengths. It is also possible to alter such polyketide structures by deleting the KR gene.3 Lastly, a considerable degree of variability exists for non-acetate starter units among certain naturally occurring aromatic PKSs; this might offer another opportunity for obtaining novel polyketides via genetic engineering. Use of an expression system such as that described here, which generates the products of PKSs, without post-PKS modification reactions, may have a particular advantage in producing targets for subsequent synthetic or enzyme-catalyzed derivatization.
Experimental Section Bacterial Strains, Plasmids, and Culture Conditions. S.coeficolor CH999 was used as a host for transformation by all plasmids. The derivation of this strain was described elsewhere.) DNA manipulations were performed in Escherichia coli MC1061. Plasmids were passaged through E. coli ET12567 (dam dcm hsdSCm1)l6togenerateunmethylated DNA before their use to transform 5'. coelicolor. E. coli strains were grown under standard conditions. S. coelicolor strains were grown on R2YE agar plates" rather than in liquid media because of the apparently more abundant production of metabolites on agar media. Manipulation of DNA and Organisms. Polymerase chain reaction (PCR) was performedusing Taq polymerase (Perkin Elmer Cetus) under conditions recommended by the enzyme manufacturer. Standard in vitro techniques were used for DNA manipu1ations.l8 E. coli was transformed with a Bio-Rad E. Coli pulsing apparatus using protocols provided by Bio-Rad. S. coelicolor was transformed by the standard procedure," and transformants were selected using 2 mL thiostrepton overlay. Comtructioa of Plasmids Containing Recombinnnt PKSs. All plasmids are derivatives of pRM5.) Fren PKS genes were amplified via PCR with S'and 3'restriction sites flanking thegenes inaccordancewith thelocation of cloning siteson pRM5 (i.e. Pad-NsiI for ORFl, NsiI-XbaI for ORF2, and XbaI-PstI for ORF3). Following subcloning and sequencing, the amplified fragments were cloned in place of the corresponding fragments in pRM5 to generate the plasmids for transformation. Production and Purification of Polyketides. For initial screening, all strains were grown at 30 OC as confluent lawns on 10-20 plates each containing approximately 30 mL of agar medium for 6-8 days. Additional plates were made as needed to obtain sufficient material for complete characterization. CH999 was a negative control when screening for potential polyketides. The agar was finely chopped and extracted with ethyl acetate/ 1% acetic acid. The concentrated extract was then flashed through a silica gel (Baker 40 pm) chromatography column in ethyl acetate/ 1%acetic acid. The primary yellow fraction was further purified via high-performance liquid chromatography (HPLC) using a 2040% acetonitrile/water/ 1%acetic acid gradient on a preparative reverse-phase
(16) MacNeil, D. J. J. Baeteriol. 1988, 170, 5607. (17) Hopwood, D. A.; Bibb, M. J.; Chater, K. F.; Kieser, T.; Bruton, C. J.; Kieser, H. M.; Lydiate, D. J.; Smith, C. P.; Ward, J. M.; Schrempf, H. Genetic Monipulation of Streptomyces. A Loboratory Manual; The John Innes Foundation: Norwich, U.K., 1985. (18) Maniatis, T.; Fritsch, E. F.; Sambrook, J. Molecular Cloning: A Loboratory Manual; Cold Spring Harbor Laboratory: New York, 1982.
J . Am. Chem. Soc.. Vol. 1 1 5, No. 25, 1993
11675
(C-18) column (Beckman). Absorbance was monitored at 280 and 410 nm. The compounds reported in this paper are the major polyketide products obtained with the above purification procedure. Several other minor compounds were present but no effort was made to characterize them. In general, the yield of purified product from these strains was approximately 10mg/L for compounds6 and 7 and 5 mg/L for compounds 4 and 5. [1,2'~2]AcetateFeeding Experiments. Two 2-L flasks each containing 400 mL of modified NMP mediumlg were inoculated with spores of S. coelicolor CH999/pRM18 and incubated in a shaker at 30 OC and 300 rpm. Toeachflaskwasadded 50mgofsodium [1,2-13Cz]acetate(Aldrich) at 72 and 96 h. After 120 h, the cultures were pooled and extracted with two 500-mL volumes of ethyl acetate/ 1% acetic acid. The organic phase was kept, and purification proceeded as described above. Approximately 2mgofRM18 wasobtainedin thisexperiment. I3CNMRdata(16 512 scans) indicate approximately a 2-3% enrichment (estimated by comparing peak areas to the natural abundance "C peak area). NMR Spectroscopy. All spectra were recorded on a Varian XL-400 except for HMBC analysis of RM18 (6), which was performed on a Nicolet NT-360. I3C spectra were acquired with continuous broadband proton decoupling. For NOE studies of RM18 (6), the 1D difference method was employed. All compounds were dissolved in DMSO-& (Sigma, 99+ atom % D), and spectra were referenced internally to the solvent. Hydroxyl resonances were identified by adding D20 (Aldrich, 99 atom % D) and checking for disappearance of signal. 3,8-Dihydroxy-l-methylanthraquinone-2-carboxylic Acid (4): IH NMR (DMSO-& 400 MHz) 6 12.86 (s, lOH), 7.73 (dd, J = 8.1, 7.6, Hz,lH),7.64(dd,J=7.6,1.2Hz,lH),7.61(s,lH),7.34(dd,J=8.3, 1.2 Hz, lH), 2.69 ( 8 , 3H); I3C NMR (DMSO-&, 100 MHz) 6 189.4, 182.2,168.5,161.4,158.9,140.8,136.3,136.0,132.5,131.4,124.4,122.5, 118.3, 116.8, 112.3, 19.9; FD-MS m / e 298. A l o e ~ a p o ~ r II i n (5): 'H NMR (DMSO-&, 400 MHz) 6 13.00 (8,
lOH), 7.72 (t, J = 8.0 Hz, lH), 7.63 (dd, J = 7.5, 1.2 Hz, lH), 7.47 (d, J = 2.6 Hz, lH), 7.34 (dd, J = 8.3, 1.2 Hz, lH), 7.06 (d, J = 2.6 Hz, lH), 2.72 (s, 3H); "C NMR (DMSO-&, 100 MHz) 6 189,1,182.5, 163.4,161.5,145.4,137.0,135.9,132.7,124.9,124.3,123.0,118.2,116.6, 112.5, 23.7; FD-MS m / e 282.
Acknowledgment. We are very grateful to Maureen J. Bibb and David H. Sherman for providing fren clones and sequence information prior to publication, and to Mervyn Bibb, Keith Chater, Tobias Kieser, Peter Revill,and Tom Simpson for valuable comments on the manuscript. We also thank Francisco Malpartida for act DNA and sequenceinformation and Virginia Miner (Acorn NMR) for assistance with HMBC analysis. This research was supported in part by grants from the National Science Foundation (BCS-9209901) and the American Cancer Society (IRG-32-34) to C.K. Acknowledgement is also made to the donors of the Petroleum Research Fund, administered by the American Chemical Society, and to the Camille and Henry Dreyfus Foundation for partial support of this research in C.K.'s laboratory. D.A.H. acknowledges financial support from the Agricultural and Food Research Council and the John Innes Foundation. (19) Strauch, E.; Takano, E.; Baylis, H. A.; Bibb, M. J. Mol. Microbiol. 1991, 5, 289.
